Prediction of Seismic Wave Velocity Using Artificial Intelligence

Abstract

AbstractIn order to determine the dynamic properties through wave velocities, the correlation between wave velocity and other properties such as Standard Penetration Test value, bulk density, and depth has been investigated in the present study. The seismic refraction method fails to detect a weak soil (low density) layer due to its working principle (Snell’s law). The primary focus of the study is to develop an artificial neural network (ANN) model for quick velocity prediction as well as to explore its capability to detect the weak layer. A new zone-specific parameter was introduced for the model to learn zones having a weak soil layer. Four sites located inside the campus were selected for the preparation of the dataset for training. The robustness of the model was tested by training the proposed ANN architecture with a published dataset of Lucknow City. The ANN gave better results upon comparison with regular statistical models. Moreover, the ANN was able to predict the velocity of the weak soil layer with decent accuracy and also provided a satisfactory prediction of the velocity profile of Lucknow City. Other geophysical tests suffering from similar shortcomings can be overcome using a data learning approach.KeywordsSeismic refraction testSPTShear wave velocityPenetration resistanceArtificial neural networkArtificial intelligence